Consider a case where you need to manage sensitivity labels in semantic models and lakehouses. How would you ensure that these labels are consistently applied across different data sources and formats? Describe the strategies you would use to maintain consistency and accuracy.